High Density Individually Addressable Nanowire Arrays RecordIntracellular Activity from Primary Rodent and Human Stem CellDerived NeuronsRen Liu,† Renjie Chen,† Ahmed T. Elthakeb,† Sang Heon Lee,† Sandy Hinckley,¶ Massoud L. Khraiche,†

John Scott,‡ Deborah Pre,¶ Yoontae Hwang,† Atsunori Tanaka,§ Yun Goo Ro,† Albert K. Matsushita,§

Xing Dai,†,# Cesare Soci,# Steven Biesmans,¶ Anthony James,∥ John Nogan,∥ Katherine L. Jungjohann,∥

Douglas V. Pete,∥ Denise B. Webb,∥ Yimin Zou,‡ Anne G. Bang,¶ and Shadi A. Dayeh*,†,§,⊥

†Integrated Electronics and Biointerfaces Laboratory, Department of Electrical and Computer Engineering, ‡Neurobiology Section,Biological Sciences Division, §Graduate Program of Materials Science and Engineering, and ⊥Department of NanoEngineering,University of California San Diego, 9500 Gilman Drive, La Jolla, California 92093, United States¶Conrad Prebys Center for Chemical Genomics, Sanford Burnham Prebys Medical Discovery Institute, 10901 North Torrey PinesRoad, La Jolla, California 92037, United States#Division of Physics and Applied Physics, Nanyang Technological University, 21 Nanyang Link, Singapore 637371, Singapore∥Center for Integrated Nanotechnologies, Sandia National Laboratories, Albuquerque, New Mexico 87185, United States

*S Supporting Information

ABSTRACT: We report a new hybrid integration schemethat offers for the first time a nanowire-on-lead approach,which enables independent electrical addressability, is scalable,and has superior spatial resolution in vertical nanowire arrays.The fabrication of these nanowire arrays is demonstrated to bescalable down to submicrometer site-to-site spacing and can becombined with standard integrated circuit fabrication tech-nologies. We utilize these arrays to perform electrophysio-logical recordings from mouse and rat primary neurons andhuman induced pluripotent stem cell (hiPSC)-derivedneurons, which revealed high signal-to-noise ratios andsensitivity to subthreshold postsynaptic potentials (PSPs).We measured electrical activity from rodent neurons from 8 days in vitro (DIV) to 14 DIV and from hiPSC-derived neurons at 6weeks in vitro post culture with signal amplitudes up to 99 mV. Overall, our platform paves the way for longitudinalelectrophysiological experiments on synaptic activity in human iPSC based disease models of neuronal networks, critical forunderstanding the mechanisms of neurological diseases and for developing drugs to treat them.

KEYWORDS: Nanowire, neuron, intracellular, extracellular, subthreshold, drug-screening

Nanowire geometries are ideal for interfacing with cells andmeasuring intracellular potentials of neurons with

minimal invasiveness.1−6 Prior works have demonstrated singlenanowire devices7,8 or devices encompassing ensembles ofseveral nanowires,9−13 but individual electrical addressability ofa single nanowire in a vertically standing array of nanowires,which is important for localizing the origin of action potentials,has not been accomplished before. Additionally, interfacingwith human neurons, important for drug screening, is yet to bedemonstrated. Here, we report a new hybrid integrationscheme that offers for the first time a nanowire-on-leadapproach, which enables independent electrical addressability,is scalable, has a high spatial resolution, and has a sensitivitythat allows the detection of subthreshold PSPs. The fabricationof these nanowire arrays is demonstrated to be scalable down tosubmicrometer site-to-site spacing and can be combined with

standard integrated circuit fabrication technologies. Physio-logical recordings from rodent primary neurons and humaninduced pluripotent stem cell (hiPSC)-derived neuronsrevealed high signal-to-noise ratios (SNRs) for both positiveand negative measured potentials. We measured electricalactivity from rodent neurons from 8 DIV to 14 DIV, and fromhiPSC-derived neurons at 6 weeks in vitro post culture, andfound intimate nanowire/neuron interaction validated bytransmission electron microscopy (TEM). The techniquecontrasts to the standard patch-clamp,14−17 which is destructiveand unscalable to large neuronal densities and to long recordingtimes,18 or to planar multielectrode arrays that enable long-

Received: November 13, 2016Revised: March 24, 2017Published: April 6, 2017

Letter

pubs.acs.org/NanoLett

© 2017 American Chemical Society 2757 DOI: 10.1021/acs.nanolett.6b04752Nano Lett. 2017, 17, 2757−2764

term recordings, but can just measure extracellular potentialsand lack the sensitivity to subthreshold potentials.19

To measure minute potential changes in individual cells athigh spatial resolution in neuronal networks, it is important todevelop an integration scheme for high-density individuallyelectrically addressable out-of-plane nanowire neuronal probes.To achieve very high densities of individual nanowires that aresuitable for mapping individual units in neuronal networks, wedevised a novel all solid-state wafer bonding integration schemeon patterned Ni electrical contacts and leads (Figure 1, FiguresS1 and S2)20 leading to a superior high-density nanowire−neuron interface platform. This bonding scheme is essentialbecause conventional low temperature eutectic bonding doesnot provide the lead-to-lead electrical isolation necessary forindividual electrical addressability of single sites in a high-density nanowire array.21 Instead, we utilize the thermallydriven solid-state diffusion of Ni into Si at a low temperature(400 °C), traditionally used to make self-aligned contacts fortransistors in the semiconductor industry,22 to bond Sisubstrates to optically transparent and electrically insulatingsapphire substrates that were predefined with Ni patterns. Bydoing so, we achieve two goals with the Ni layer: (1) bondingand fusion of a thin (∼50 μm) Si substrate to the underlyinghost substrate and (2) embedding electrical leads underneathactive or passive Si components in the bonded substrate withlow contact resistance (Figure S1). The integration technique is

general to any other substrate that can sustain the NiSi reactiontemperature (starts at 300 °C), including complementary metaloxide semiconductor (CMOS) integrated circuits and advancedplanar23,24 and out of plane device geometries are attainablethrough this method. Additionally, the optical transparency ofsapphire enables light excitation and transmission imagingdegrees of freedom in our platform. Overall, the integrationtechnology developed in this work is the first to enableelectrical addressability for individual vertically standingnanowires registered precisely over underlying metal leads.This individual electrical addressability of nanowires haspotential to enable precise measurements of activity ofindividual units in neuronal networks, and to detect miniaturerelease of neurotransmitters, especially important for inves-tigating the synaptic properties of networks of neurons in thecontext of neurological diseases, as for the characterization ofpre- or post- synaptic defects based on amplitude or frequencymodifications in the subthreshold postsynaptic potentials.25,26

Fabrication of our in vitro platform starts with photo-lithography and e-beam lithography (EBL) patterning ofelectrode leads on a sapphire substrate, as shown schematicallyin Figure 1A-i.27 The electrode leads have metal stacking of Ti/Ni/Ti/Ni (30 nm/200 nm/50 nm/200 nm) for adhesion/conduction/diffusion-barrier/silicidation purposes, respectively.When a thin Si chip, 5 mm × 5 mm × 50 μm, is brought intocontact with the metal leads on the sapphire substrate, a

Figure 1. (A) Illustration of fabrication procedure for high density electrically isolated nanowire probes by solid-state wafer bonding. (i) Metal stackwith Ni topmost layer is patterned by a combination of photolithography and electron beam lithography atop an electrically insulating andtransparent sapphire substrate. (ii) Si is bonded to the substrate in i by nickel silicidation. (iii) Si wafer is thinned down to the desired wire height. Nimasks are then defined by electron beam lithography and aligned to the bottom plane Ni pattern. (iv) Si nanowires are etched by an SF6/C4F8plasma etch step. (v) SiO2 PECVD is then deposited and is selectively etched to expose the tips of the Si nanowires. (B, C) SEM images of an 8 × 8Si nanowire array (B) after etching and (C) after SiO2 passivation. Scale bar in B is 5 μm and in C is 3 μm. (D, E). Energy-dispersive X-rayspectroscopy (EDX) of the (D) oxygen signature of the SiO2 passivation layer, (E) nickel for etching mask (top), NiSi region, and conducting lowermost layer, and (F) Ti as the diffusion barrier (top) and adhesion layer (bottom). Scale bars are 1 μm. (G) High magnification TEM image of theNiSi/Ti/Ni/Ti underneath the Si nanowire highlighting the interfacial structure between the bottom conducting lead and the Si nanowire. Scale baris 200 nm. The bottom panels are HRTEM images at the interface between Si and NiSi with electron beam axis aligned in the Si [1 10] (left panel),and the NiSi [1 10] Zone axis in (right panel) to display the crystalline interface. Scale bars in bottom panels are 2 nm.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.6b04752Nano Lett. 2017, 17, 2757−2764

2758

moderate heat treatment (400 °C) and compressive pressure(around 10 MPa) in a vacuum chamber with forming gas (H23−3.8% in N2) flow initiates a diffusion reaction between Niand Si. The formed NiSi alloy fuses the two substrates together(Figure 1A-ii). The temperature 400 °C was chosen because itled to reacted NiSi leads that formed ohmic-like contacts withthe heavily doped Si (Figures S1 and S2). The height of thedesired nanowire array can be adjusted by an SF6 inductivelycoupled plasma (ICP) and reactive ion etch (RIE) that thinsthe Si substrate to around 8−10 μm (Figure 1A-iii). Anotherelectron-beam lithography step patterns Ni etch mask dotsaligned exactly at the tips of the Ni leads underneath the Sisubstrate (Figure 1A-iii). An SF6/C4F8 based ICP/RIE processis used to etch the Si everywhere except regions masked by theNi dots, leaving vertically standing Si nanowires on NiSi leads(Figure 1A-iv and Figure 1B), which have diameters in the

range of 100−200 nm in this work. To prevent electrochemicalreactions anywhere except at the nanowire tips, the entire chipis coated by SiO2, which is then selectively etched at the contactpads and nanowire tips (Figure 1A-v, Figure 1C, Figure S4).The scanning electron microscope (SEM) images of the Sinanowire arrays in Figure 1B and C demonstrate a high packingdensity of 6.25Million/cm2 at a pitch of 4 μm. We alsodemonstrated submicrometer pitch at a site-to-site spacing of750 nm (Figure S3). The array geometry can be tailored for theoptimal growth of neuronal networks28 that are interconnectedwith sealed microfluidic channels that can allow growth ofneurites and synaptic connections but prevent cell-body platinginside the channels (Figures S5 and S6). The electrochemicalimpedance in all of these configurations is relatively uniform(Figure S5 and S6) and validated a capacitively dominantcoupling behavior (phase in Figure S7) with neuronal activity.

Figure 2. Modeling/characterization/measurement of the electro-neural system on mouse hippocampal neurons. (A) Electrical circuit models andcorresponding SEM images of the electrode-cell engulfment. (i) Intracellular electrode configuration. (ii) Extracellular electrode configuration. Theelectrical circuit model of the electro-neural system starts by the cell culture (yellow) interfacing with the electrode, all the way to the read-outelectronics represented by the amplifier block (green). INa and IK represents the Na+ and K+ ionic currents, and RNa and RK are the correspondingion channel resistances. Rseal is the seal resistance at the cell-nanowire interface. REC represents the resistance of the electrochemical reaction at theelectrode tip, and CEC is the double layer capacitance. Cstray and Camp are the bare electrode wire and amplifier’s input parasitic capacitances,respectively. (B) Spontaneous action potentials recorded on mouse hippocampal neurons. (i) Recordings showing the positive measured signal (∼20mV p-p) (left in blue) and after deconvolution (right in red). (ii) Recordings showing the negative measured signal (∼10 mV p-p). Small potentialfluctuations were captured in both cases. (C) Effect of external application of glutamate/TTX to the recordingsolution. (i) Baseline recordingsshowing the normal cell firing activity. (ii) After addition of 196 μM Glutamate to the recording solution, increased activity is observed. A sampleaction potential is shown on the right (blue), and its deconvoluted potential is shown below (red). (iii) After the application of 1.5 μM TTX, the cellactivity is blocked. (D) Effect of addition of 13.2 mM KCl to the recording solution (i) Baseline recording showing very quiet cell activity. (ii)Following the application of KCl, increased cell activity was observed with a train of action potentials. Similarly, sample action potential is shown onthe left (blue), and its deconvolution is shown below (red). Electrophysiology was performed at 8 DIV.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.6b04752Nano Lett. 2017, 17, 2757−2764

2759

Transmission electron microscopy (TEM) and elementalmapping by energy-dispersive X-ray spectroscopy (EDX) of theSi nanowires demonstrate crystalline structures and interfacesand highlight the usefulness of each layer: Si constitutes themain body of the sensor, SiO2 constitutes the passivationoutermost cylinder around the bottom portion of the nanowire(Figure 1D), Ni is used for silicidation bonding and as a currentconduction layer (Figure 1E top and bottom, respectively), andTi is used as a Ni diffusion barrier and adhesion layer (Figure1F top and bottom, respectively). A zoomed-in TEM image atthe bottom of the wire in Figure 1G highlights the interfaciallayers crucial for the free-standing and individually electricallyaddressable nature of the Si nanowire. The high-resolutionTEM (HRTEM) images in the lower panels of Figure 1Gillustrate the crystalline nature of the interface between NiSiand Si at the bottom of the wire (see also Figure S9). Similarbonding structure, morphology, and interfaces were validatedon SiO2/Si substrates (Figure S10) illustrating the versatility ofour bonding scheme.We next investigated the biological interfaces established

between our Si nanowires and neurons and the resultingrecorded electrophysiological activity. We packaged our deviceson commercial electrophysiology printed circuit boards (FigureS8) and tested their feasibility for electrophysiology andpharmacology using rodent hippocampal primary neuronsand human iPSC-derived neurons (Supporting Informationsections 3 and 4). For both rodent primary and hiPSC-derivedneurons, we find strong interaction between neurons andnanowires characterized by cell outgrowth and engulfment tothe vertical Si nanowires (Figure 2A and Figure 4A, primarymouse neurons; Figure S12 uncolored SEM images). The cell−electrode interface is generally established in intracellular and

extracellular configurations both of which can provide sufficientcoupling between the cell and the electrode to enable highfidelity recordings.The amplitude and shape of recorded potentials are governed

by the electrochemical interface at the surface of the nanowireand the degree of sealing for the cell membrane to the nanowireitself, and by the measurement system.29−31 To account forthese effects in both intracellular and extracellular config-urations (Figure 2A), we developed a circuit model based onexperimental measurements for each component of theelectrode/neuron interface (see Supporting Informationsection 6). Results using analytic transfer functions, circuitsimulations, and empirical transfer functions for the overallmeasurement system provided excellent agreement with themeasured potentials.One of the main signatures of our model is the use of current

source and different conductance for the Na+ and K+ channelsthat is commensurate to the physical origin of the fasterdepolarization and repolarization in action potentials. Thesewere calibrated to patch-clamp measured inward and outwardionic currents from similar cultured cells (Figure S15a). Thelonger duration of the patch-clamp action potentials has beenpreviously related to maturity, culture conditions, and exacttemperature of the cell culture for both hiPSC32 and rodentcortical neurons,33 which was also observed in our referencepatch-clamp measurements (Figure S15a). Our electrophysio-logical models using either analog circuit analysis or empiricalsimulations based on detailed electrochemical characterization,similar to the models of Spira et al.29−31 are self-consistentmodels that can accurately reproduce both the time and theamplitude of deconvoluted neuronal signals measured fromnanowires.

Figure 3. Recording of a 99 mV action potential and pharmacological experiments for validating subthreshold potentials at 10 DIV. (A) Spontaneousactivity measured on channel 44 showing subthreshold oscillations that are illustrated in the insets of a 1 s time window at the beginning of the trace(lower inset) and just at the action potential generation (top inset). (B) Concurrent recording with A from a nearby channel 46 showing somepotential oscillations and negative spikes that occur simultaneously with those in A. (C) Recording with a solution containing D-APV (50 μM),CNQX (10 μM), and PTX (1 mM) to block NMDA, AMPA, and GABA receptors, respectively, shows a single oscillation prior to the spike on thesame channel as in A. (D) Same in C from channel 37 showing no oscillations prior to the large action potential.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.6b04752Nano Lett. 2017, 17, 2757−2764

2760

Pharmacological stimulation and inhibition validated thephysiological origin of the measured potentials. After glutamatewas added to the recording solution, we observed an increase ofthe cell spontaneous activity (Figure 2C-ii) with respect tofrequency and amplitude when compared to the baseline(Figure 2C-i) measured on the same channel. The bathapplication of tetrodotoxin (TTX) inhibited the activity on thesame channel (Figure 2C-iii). Similarly, for the extracellularconfiguration, the bath application of KCl led to increasedactivity (Figure 2D-ii) relative to the baseline recording (Figure2D-i), which was also eliminated by TTX treatment.Strikingly, physiological measurements on mouse hippo-

campal neurons cultured for 10−13 days on our platformdisplayed small potential fluctuations prior to the positive(Figure 2B-i) and negative (Figure 2B-ii) firing events. TheSNR of these prespike potentials is 20-times as evident in theinset of Figure 2B-i, and their shape is clearly different fromcoupled action potentials from different cells or channels. Thelargest action potentials that we measured 10 DIV is 99 mV, asshown in Figure 3A, which demonstrates for the first time thatnanowires can measure intracellular potentials with similarmagnitudes to that of patch-clamp. We generally observedoscillations prior to spikes as highlighted in the insets of Figure3A. For a simultaneously measured extracellular potential fromanother channel (Figure 3B) that is 5.65 μm apart from the

intracellular channel, we also observed potential oscillationsprior to the extracellular spike. Upon inspecting all otherchannels, we did not observe action potentials that might beelectronically coupled to these two channels. We thenembarked on validating that these potential oscillations aresubthreshold synaptic potentials. To do so, we pharmacolog-ically blocked both excitatory and inhibitory receptors byadding D-APV to block NMDA receptors, CNQX for blockingAMPA receptors, and Picrotoxin for preventing the binding ofthe inhibitory neurotransmitter GABA to its receptors (seeSupporting Information section 4 (1)). After adding theblockers, we observed insignificant number of small prespikeoscillations (Figure 3C) or no oscillations (Figure 3D), whichsuggested that our system has the sensitivity to detect miniaturerelease of neurotransmitters at a quantal level. Given that ournanowires can sometimes measure intracellular potentials withan SNR of 1700 (Figure 3A), it is not surprising that they canalso resolve subthreshold activity that we validate with standardpharmacological experiments here. This can be attributed to theheight of our nanowires, which is >6.5 μm compared to theshorter than 2 μm nanowires in prior works, providing larger Sisurface interaction area with adherent neurons.The development of high-throughput, subcellular neuro-

technologies has the potential for application to drug screeningon neurological disease models. We therefore tested the

Figure 4. Recording from hiPSC-derived cortical neurons. (A) Colorized angle-view SEM image showing cell morphology and neurite outgrowth tonearby nanowires. Scale bar is 4 μm. (B) Measured potentials on channels 6−8 showing positive potentials for wires 7 and 8 inside the cell andnegative potential for wire 6 outside the cell. (C) Colorized SEM image of the Pt coated and FIB thinned cross-section of the cell in panel B. Theimage shows the cell (yellow colored) engulfing the tops of nanowires 7 and 8, and portions of cell membrane around nanowire 6. Scale bar is 2 μm.Electrophysiology was performed 6 weeks post in vitro culture. (D, E) Higher magnification colorized TEM images collected and stacked to providea higher resolution close-up to the nanowire-cell interface. The slight bending of the nanowires is likely to have happened during neuron dehydrationin preparation for SEM imaging (section 5 of Supporting Information). Scale bars are 200 nm. (F) Recordings on nanowire number 7 illustratingconsistent oscillations prior to action potential firing at different recording times.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.6b04752Nano Lett. 2017, 17, 2757−2764

2761

sensitivity of our platform with two clinically relevant subtypesof human iPSC-derived neurons, cortical and dopaminergic, forwhich we had demonstrated electrical activity on conventionalmicroelectrode array recordings (Figure S11).Human iPSC-derived cortical neurons cultured on our

platform overlapped with multiple nanowires (Figure 4A).Post preparation for SEM on this platform, evidence of neuriteoutgrowth to nearby nanowires, is also apparent in Figure 4A.Nanowires 7 and 8 displayed positive action potentials, whereasnanowire 6 displayed a negative action potential (Figure 4B).From the SEM image of Figure 4A, we can note theextracellular nature of the interface with nanowire 6. Touncover the nature of the nanowire/neuron interaction fornanowires 7 and 8, we performed a sequential focused ionbeam (FIB) cut and thinning of a 300−400 nm slice on thesample in regions of wires marked 6, 7, 8 (Figures S17−S21)post Pt plating. The sample is thin enough to allow electrontransmission for TEM characterization without riskingsignificant damage to the cell body during the FIB millingprocess. Figure 4C shows an SEM image of the FIB sliceshowing a clear dark contrast of the cell around nanowires 7and 8. The TEM images in Figures 3D and 4E along thesubstrate-nanowire-cell regions demonstrated that the nano-wires displayed intimate interaction with neurons that is notaided by tension due to neuron adhesion and spreading on thesubstrate surface or peptide-modification,34−36 nor assisted withthe highly invasive electroporation.9,10 The continuity of theinclined interface below the SiO2 passivation layer as seen tothe right of the nanowires in Figure 4D and E is suggestive ofintracellular penetration of the nanowires into the humancortical neuron cell body, which is commensurate with thepositive potentials measured with wires 7 and 8.37 However, itis also possible that the cell fixation and FIB thinning can leadto artifacts in the observed interface. Also in these recordings,we observed the prespike potentials with sharp rise and slowdecay times, which are similar in shape to the excitatorypostsynaptic potentials (EPSPs) observed by Hai et al. using Aumushroom electrodes.38 It is possible however that the positivepotentials are measured in a juxtacellular configuration due totheir smaller amplitude than those observed in Figures 2, 3, and

5 since we did not validate their nature by blocking thepostsynaptic receptors. We note that out of the two hiPSC cellson top of the nanowire array (Figure S17), we measuredpositive potentials from one neuron. The individuallyaddressable nanowires can record multiple positive potentialsfrom a single neuron.Dopaminergic neurons are a clinically relevant cell type for

study of neurodegenerative disease and neuropsychiatricdisorders. We determined whether hiPSC-derived dopaminer-gic neurons could also survive and demonstrate physiologicalfunction on our nanowire device. Six weeks postplating hiPSC-derived dopaminergic neurons on the nanowire platform, weobserved distinctive single slope rise potentials compared to themultioscillation behavior exhibited by the mouse hippocampalneurons and hiPSC-derived cortical neurons prior tospontaneous action potential firing for both positive andnegative polarities (inset, Figure 5A). On some channels, weobserved small potential oscillations with varying frequency andamplitude as shown Figure 5B. We then performedpharmacological studies after 2 weeks postastrocyte cocultureon the hiPSC-derived dopaminergic neurons (6 weeks postinitial culture). Upon bath application of KCl, the firing rateincreased and then all activity stopped after the addition ofTTX as shown in Figure 5C. In conclusion, hiPSC-derivedneurons not only survived several weeks intimately interfacedwith the nanowires, but also we observed extensive electro-physiological activity over time, which led to the possibility oflongitudinal electrophysiological experiments on synapticactivity on in vitro human neuronal networks.The electrophysiological recordings from our cultures did

not display trains of large action potentials, but just one or twoAPs, which is consistent with recordings from immatureneurons and network observed on both human and rodentcultures from other groups.32,39 Therefore, we performedimmunocytochemistry analyses on primary neurons cultured onour platform. The rat cortical neurons fixed at 15 DIV showedthat the cells are well adherent to the nanowire arrays, stainedpositive for Tuj1, a specific neuronal marker, with some limitedneurite extension. (Figure S22). The normal formation of amature neuronal network can be compromised by the height of

Figure 5. Recording from hiPSC-derived dopaminergic neurons 6 weeks post in vitro culture. (A) Spontaneous action potentials. (i) Positivepotential recordings showing the measured signal (∼30 mV p-p) (left in blue) and after deconvolution (right in red). (ii) Negative potentialrecordings (∼6 mV p-p) (left in blue) and after deconvolution (right in red). (B) Potential oscillations recorded prior to action potential firing. (C)Effect of external application of stimulating/inhibiting media. (i) Increase of action potential firing following 13.2 mM KCl bath application. (ii)Increased activity abolished after the application of 1 μM TTX.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.6b04752Nano Lett. 2017, 17, 2757−2764

2762

our nanowires that pin the cells above the planar regions of theplatform. This can be improved by creating 3D islands at thebottom of the nanowires that can promote networkconnectivity on a 3D surface. In the extreme case of shorteningthe nanowires to heights that are similar to that is reportedpreviously, the intracellular capability of the nanowires might becompromised. Increasing the spacing between the nanowiresmay also strengthen the network activity of the culture as notedin the recent work of Shmoel et al.40 but will compromise theirdensity. Further investigation of the limited neurite extension inour cell culture will lead to improved network synapticconnections and to more robust and consistent measurementsof action potentials and PSPs. (Figures 2−4).It is also important to report the our primary rodent neuron

experiments were conducted on 5 devices with successfulmeasurements for several days for both intracellular-like andextracellular-like potentials on 4 out of the 5 devices. Themeasured intracellular-like potentials varied from 0.1 mV to 99mV. The hiPSC experiments were conducted on 4 devices withsuccessful measurements for several days for both intracellular-like and extracellular-like potentials on 4 out of the 5 sets, andfor one device over several weeks (4 weeks and then 6 weeksafter culture). The measured intracellular-like potentials variedfrom 0.1 mV to 35 mV. From Figure S17, we can observe 2cells in the 28 μm × 28 μm square array of nanowires leadingto a cell density of ∼255 100 cells/cm2. One can also observefrom Figure S17 that two out of the 12 nanowires in contactwith the cultured cells resulted in intracellular-like measure-ments. As observed by other groups the variability of theAmplitude and duration of APs can be due to differentengulfment level or cleft width between the neuron and thenanowire.40 The yield on nanowire internalization to cells canalso be improved, for example with peptide modification.36

Measurement of intracellular action potentials usingindividually addressable Si nanowire probes, from both mouseand human neurons, opens new prospects on mappingneuronal activity in large networks, while the sensitivity tosubthreshold postsynaptic potentials from multiple neuronsopens new possibilities to study synaptic transmissionmechanisms and plasticity, particularly important for inves-tigating neurological diseases. Given the scalability of ourarrays, the simultaneous recording of minute changes in cellpotentials can uncover details on the synthesis, processing, andexecution of neuronal network activity. In vitro, highly paralleldrug screening experiments can be performed using the humanrelevant iPSC cell line and without the need of the laboriousnonscalable patch-clamp. In vivo, targeted modulation ofindividual neural circuits or even single cells within a networkbecomes possible, and implications for bridging or repairingnetworks in neurologically affected regions become withinreach. Overall, our platform and modified versions thereof havethe potential to lead to transformative technologies for both invivo and in vitro applications.

■ ASSOCIATED CONTENT

*S Supporting InformationThe Supporting Information is available free of charge on theACS Publications website at DOI: 10.1021/acs.nano-lett.6b04752.

Detailed information regarding fabrication, packaging,electrochemical impedance spectroscopy, TEM samplepreparation, cell culture, pharmacological stimulation and

inhibition, SEM imaging, modeling and simulation, andnanowire−cell interface characterization (PDF)

■ AUTHOR INFORMATIONCorresponding Author*E-mail: sdayeh@ece.ucsd.edu.ORCIDRenjie Chen: 0000-0002-3145-6882Shadi A. Dayeh: 0000-0002-1756-1774Author ContributionsS.A.D. conceived the integration scheme and platform design,led the project, wrote the manuscript, and received input fromall coauthors. R.L., R.C., Y.H., A.T., Y.G.R., X.D., and S.A.D.fabricated the platform with input and equipment training fromC.S., A.J., J.N., and D.B.W. R.L. performed the FIB sectioningand TEM characterization with input from D.V.P. and K.L.J.;J.S., S.H.L., and Y.Z. cultured the primary rodent neurons, andR.L., S.H.L., M.L.K., D.P., and S.A.D. performed electro-physiology on these devices. S.H., S.B., and A.G.B. performedthe hiPSC culture and MEA analyses, and S.H.L, S.H., M.L.K.,and A.K.M. performed electrophysiology on the hiPSC devices.A.T.E. and S.A.D. carried out the electrochemical interfacemodeling.NotesThe authors declare no competing financial interest.

■ ACKNOWLEDGMENTSS.A.D. acknowledges support of the National ScienceFoundation under CAREER Program No. ECCS-1351980,which funded most aspects of this work. S.A.D. acknowledgessupport for efforts in this work under NSF DMR-1503595, theCenter for Brain Activity Mapping at UC San Diego, aQualcomm Institute CSRO Award CITD137, and a LaboratoryDirected Research and Development Exploratory Research(LDRD-ER) award to S.A.D. from Los Alamos NationalLaboratory. Y.Z. acknowledges support of NIH R21MH099082, March of Dimes award, and Frontiers ofInnovation Scholars Program (FISP) to J.S. The work wasperformed in part at the Center for Integrated Nano-technologies (CINT), U.S. Department of Energy, Office ofBasic Energy Sciences User Facility at Los Alamos NationalLaboratory (Contract No. DE-AC52-06NA25396), and SandiaNational Laboratories (Contract No. DE-AC04-94AL85000).We are grateful for technical support from the nano3 cleanroom facilities at UC San Diego’s Qualcomm Institute andfrom the Integration Laboratory at CINT.

■ REFERENCES(1) Kim, W.; Ng, J. K.; Kunitake, M. E.; Conklin, B. R.; Yang, P.Interfacing silicon nanowires with mammalian cells. J. Am. Chem. Soc.2007, 129 (23), 7228−7229.(2) Piret, G.; Perez, M.-T.; Prinz, C. N. Neurite outgrowth andsynaptophysin expression of postnatal CNS neurons on GaP nanowirearrays in long-term retinal cell culture. Biomaterials 2013, 34 (4), 875−887.(3) Hanson, L.; Lin, Z. C.; Xie, C.; Cui, Y.; Cui, B. Characterizationof the cell−nanopillar interface by transmission electron microscopy.Nano Lett. 2012, 12 (11), 5815−5820.(4) Gallentoft, L.; Pettersson, L. M.; Danielsen, N.; Schouenborg, J.;Prinz, C. N.; Linsmeier, C. E. Size-dependent long-term tissueresponse to biostable nanowires in the brain. Biomaterials 2015, 42,172−183.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.6b04752Nano Lett. 2017, 17, 2757−2764

2763

(5) Patolsky, F.; Timko, B. P.; Yu, G.; Fang, Y.; Greytak, A. B.;Zheng, G.; Lieber, C. M. Detection, stimulation, and inhibition ofneuronal signals with high-density nanowire transistor arrays. Science2006, 313 (5790), 1100−1104.(6) Cohen-Karni, T.; Qing, Q.; Li, Q.; Fang, Y.; Lieber, C. M.Graphene and nanowire transistors for cellular interfaces and electricalrecording. Nano Lett. 2010, 10 (3), 1098−1102.(7) Duan, X.; Gao, R.; Xie, P.; Cohen-Karni, T.; Qing, Q.; Choe, H.S.; Tian, B.; Jiang, X.; Lieber, C. M. Intracellular recordings of actionpotentials by an extracellular nanoscale field-effect transistor. Nat.Nanotechnol. 2011, 7 (3), 174−179.(8) Tian, B.; Cohen-Karni, T.; Qing, Q.; Duan, X.; Xie, P.; Lieber, C.M. Three-dimensional, flexible nanoscale field-effect transistors aslocalized bioprobes. Science 2010, 329 (5993), 830−834.(9) Xie, C.; Lin, Z.; Hanson, L.; Cui, Y.; Cui, B. Intracellularrecording of action potentials by nanopillar electroporation. Nat.Nanotechnol. 2012, 7 (3), 185−190.(10) Robinson, J. T.; Jorgolli, M.; Shalek, A. K.; Yoon, M.-H.;Gertner, R. S.; Park, H. Vertical nanowire electrode arrays as a scalableplatform for intracellular interfacing to neuronal circuits. Nat.Nanotechnol. 2012, 7 (3), 180−184.(11) Angle, M. R.; Cui, B.; Melosh, N. A. Nanotechnology andneurophysiology. Curr. Opin. Neurobiol. 2015, 32, 132−140.(12) Duan, X.; Fu, T.-M.; Liu, J.; Lieber, C. M. Nanoelectronics-biology frontier: From nanoscopic probes for action potentialrecording in live cells to three-dimensional cyborg tissues. NanoToday 2013, 8 (4), 351−373.(13) Robinson, J. T.; Jorgolli, M.; Park, H. Front. Neural Circuits2013, 7, 38.(14) Neher, E.; Sakmann, B.; Steinbach, J. H. The extracellular patchclamp: a method for resolving currents through individual openchannels in biological membranes. Pfluegers Arch. 1978, 375 (2), 219−228.(15) Martina, M.; Vida, I.; Jonas, P. Distal initiation and activepropagation of action potentials in interneuron dendrites. Science2000, 287 (5451), 295−300.(16) Henze, D.; Buzsaki, G. Action potential threshold ofhippocampal pyramidal cells in vivo is increased by recent spikingactivity. Neuroscience 2001, 105 (1), 121−130.(17) Scanziani, M.; Hausser, M. Electrophysiology in the age of light.Nature 2009, 461 (7266), 930−939.(18) Lin, Z. C.; Cui, B. Nanowire transistors: room for manoeuvre.Nat. Nanotechnol. 2014, 9 (2), 94−96.(19) Spira, M. E.; Hai, A. Multi-electrode array technologies forneuroscience and cardiology. Nat. Nanotechnol. 2013, 8 (2), 83−94.(20) Dayeh, S. A.; Chen, R.; Lee, S. H.; Liu, R.; Ro, Y. G.; Tanaka, A.Novel Integration Scheme for High-Density 3D Intracellular Probes withSubmicron Pitch and Individual Electrical Addressability, United StatesProvisional Patent, 2015.(21) Wolffenbuttel, R.; Wise, K. Low-temperature silicon wafer-to-wafer bonding using gold at eutectic temperature. Sens. Actuators, A1994, 43 (1−3), 223−229.(22) Tang, W.; Nguyen, B.-M.; Chen, R.; Dayeh, S. A. Solid-statereaction of nickel silicide and germanide contacts to semiconductornanochannels. Semicond. Sci. Technol. 2014, 29 (5), 054004.(23) Dai, X.; Nguyen, B. M.; Hwang, Y.; Soci, C.; Dayeh, S. A. Novelheterogeneous integration technology of III−V layers and InGaAsfinFETs to silicon. Adv. Funct. Mater. 2014, 24 (28), 4420−4426.(24) Chen, R.; Dayeh, S. A. Size and Orientation Effects on theKinetics and Structure of Nickelide Contacts to InGaAs Fin Structures.Nano Lett. 2015, 15 (6), 3770−3779.(25) Engelman, H. S.; MacDermott, A. B. Presynaptic ionotropicreceptors and control of transmitter release. Nat. Rev. Neurosci. 2004, 5(2), 135−145.(26) Kerchner, G. A.; Nicoll, R. A. Silent synapses and the emergenceof a postsynaptic mechanism for LTP. Nat. Rev. Neurosci. 2008, 9 (11),813−825.(27) Photolithography is used here to define the outer largeelectrodes, and the center electrodes in nanoscale are patterned by

EBL registered to the photolithography pattern utilizing a 100 kV e-beam writer (JEOL JBX- 6300FS) with beam size of ∼10 nm.(28) Merz, M.; Fromherz, P. Silicon chip interfaced with ageometrically defined net of snail neurons. Adv. Funct. Mater. 2005,15 (5), 739−744.(29) Fendyur, A.; Mazurski, N.; Shappir, J.; Spira, M. E. Formation ofessential ultrastructural interface between cultured hippocampal cellsand gold mushroom-shaped MEA-toward “IN-CELL” recordings fromvertebrate neurons. Front. Neuroeng. 2011, 4, 14.(30) Hai, A.; Spira, M. E. On-chip electroporation, membrane repairdynamics and transient in-cell recordings by arrays of gold mushroom-shaped microelectrodes. Lab Chip 2012, 12 (16), 2865−2873.(31) Ojovan, S. M.; Rabieh, N.; Shmoel, N.; Erez, H.; Maydan, E.;Cohen, A.; Spira, M. E. A feasibility study of multi-site, intracellularrecordings from mammalian neurons by extracellular gold mushroom-shaped microelectrodes. Sci. Rep. 2015, 5, 14100.(32) Pre, D.; Nestor, M. W.; Sproul, A. A.; Jacob, S.; Koppensteiner,P.; Chinchalongporn, V.; Zimmer, M.; Yamamoto, A.; Noggle, S. A.;Arancio, O. A time course analysis of the electrophysiologicalproperties of neurons differentiated from human induced pluripotentstem cells (iPSCs). PLoS One 2014, 9 (7), e103418.(33) Lee, J.; Callaway, J. C.; Foehring, R. C. Effects of temperature oncalcium transients and Ca2+-dependent afterhyperpolarizations inneocortical pyramidal neurons. J. Neurophysiol. 2004, 93 (4), 2012−2020.(34) Xie, X.; Xu, A. M.; Angle, M. R.; Tayebi, N.; Verma, P.; Melosh,N. A. Mechanical model of vertical nanowire cell penetration. NanoLett. 2013, 13 (12), 6002−6008.(35) Xu, A. M.; Aalipour, A.; Leal-Ortiz, S.; Mekhdjian, A. H.; Xie, X.;Dunn, A. R.; Garner, C. C.; Melosh, N. A. Quantification of nanowirepenetration into living cells. Nat. Commun. 2014, 5, 5.(36) Lee, J.-H.; Zhang, A.; You, S. S.; Lieber, C. M. Spontaneousinternalization of cell penetrating peptide-modified nanowires intoprimary neurons. Nano Lett. 2016, 16 (2), 1509−1513.(37) The capacitive nature of the bare Si nanowire−cytoplasminterface precludes the measurement of cell resting potentials, butcoatings of the nanowire tips with faradiac materials such as Ag/AgClor bioreagents should make this possible.(38) Hai, A.; Shappir, J.; Spira, M. E. In-cell recordings byextracellular microelectrodes. Nat. Methods 2010, 7 (3), 200−202.(39) Pedroni, A.; Minh, D. D.; Mallamaci, A.; Cherubini, E. Front.Cell. Neurosci. 2014, 8, 44.(40) Shmoel, N.; Rabieh, N.; Ojovan, S. M.; Erez, H.; Maydan, E.;Spira, M. E. Multisite electrophysiological recordings by self-assembledloose-patch-like junctions between cultured hippocampal neurons andmushroom-shaped microelectrodes. Sci. Rep. 2016, 6, 27110.

Nano Letters Letter

DOI: 10.1021/acs.nanolett.6b04752Nano Lett. 2017, 17, 2757−2764

2764